//php echo do_shortcode('[responsivevoice_button voice="US English Male" buttontext="Listen to Post"]') ?>

GlobalFoundries’ second-quarter results strengthened the commercial case behind its latest U.S. government-backed investment in silicon photonics.

The foundry reported revenue of $1.786 billion for the quarter, up 6% year over year, while non-IFRS gross margin reached 29.9%, an increase of nearly five percentage points from the prior-year period. The standout performance came from communications infrastructure and data center, where revenue rose 62% year over year and 20% sequentially. GlobalFoundries now expects that segment to grow between 50% and 60% for the full year, up from its previous forecast in the high 30% range.

The results put the company’s recently announced letter of intent with the U.S. Department of Commerce in a different light. The proposed $300 million award is not aimed at creating a market from scratch. It’s intended to accelerate development in an area where demand is already rising quickly, customers are asking for more capacity, and GlobalFoundries is expanding both its technology roadmap and its manufacturing footprint.

GlobalFoundries CEO Tim Breen said the company now expects silicon photonics revenue to more than double in 2026. It’s engaged with four of the five largest optical transceiver suppliers, has seven active customer engagements around its SCALE platform, has taped out one SCALE-related design win in the second quarter, and expects another tape-out in the third.

Demand tightens key technology corridors

Overall fab utilization was in the high 80% range, leaving the company some room to grow within its existing footprint. But the pressure is not evenly distributed. Demand is strongest in particular technology corridors tied to optical networking, especially silicon photonics and silicon germanium.

GlobalFoundries said silicon germanium capacity is already oversubscribed through 2027, while demand for silicon photonics continues to rise. The company is expanding capacity in Vermont, increasing output in Malta, New York, and qualifying additional 300-mm silicon germanium capacity in Singapore.

GlobalFoundries is not saying every fab is full, but that capacity is tightening fastest in technologies tied to AI data center connectivity, even as Breen said the company can add output relatively quickly within existing facilities—a confidence that contributed to its raised full-year forecast.

The result is a layered infrastructure strategy. In the near term, GlobalFoundries can increase output from existing fabs. Over the next several years, however, it must add more specialized capabilities in materials, fiber attachment, advanced packaging, hybrid bonding, and optical-module integration.

Photonics moves closer to the processor

The growth reflects a broader shift in how data moves through AI systems.

Copper remains widely used for short links, but bandwidth and power requirements are pushing optical connectivity progressively closer to processors. Breen said GlobalFoundries expects optical links to account for at least 70% of data center connections by 2030, adding that current evidence suggests that estimate may prove conservative.

Gregg Bartlett, GlobalFoundries’ CTO, said the company began investing in silicon photonics in 2015 because it believed copper would eventually run into bandwidth limits.

“If you believe in physics, then you should believe that photonics is ultimately going to succeed,” Bartlett told EE Times. “It just takes patience to be able to keep the sustained level of investment.”

AI has accelerated that transition. As systems connect more GPUs and other processors, the challenge is no longer only computation. It’s also moving data between processors, memory, switches, and racks without consuming excessive power or occupying too much physical space.

GlobalFoundries’ current silicon photonics business is still dominated by pluggable optical transceivers. But the company is preparing for near-packaged optics (NPO), and co-packaged optics (CPO), which move optical engines closer to the switch or processor.

GlobalFoundries expects NPO ramps to begin in 2027 and CPO ramps in 2028, with many customers pursuing both architectures and using NPO as a step toward CPO.

Capturing more of the module value

The strategic significance of SCALE is that GlobalFoundries does not intend to stop at producing silicon photonics wafers.

Bartlett described SCALE as a complete optical engine that combines a photonic integrated circuit, an electronic integrated circuit, drivers, transimpedance amplifiers, fiber attachment, bonding, and testing. The finished product would be a known-good optical module ready for integration into a larger system containing a GPU, CPU, or other processing unit.

The first generation uses thermocompression bonding. The second, supported by the proposed Commerce Department award, will move to copper hybrid bonding and higher-density integration. The program also includes next-generation fiber-attachment technologies and new modulator materials, including thin-film lithium niobate, barium titanate, indium phosphide, and polymers. GlobalFoundries expects to narrow those options based on performance, manufacturing cost, and supply chain availability.

Bartlett estimated that the wafer accounts for only 20% to 25% of the value of a completed optical module. The rest comes from integrating the photonic and electronic components, attaching fibers with low optical loss, and performing electrical and optical testing.

“We think that being able to supply an optical module is a key control point,” Bartlett said, adding that GlobalFoundries expects the module business to generate significant revenue and be “significantly margin accretive” for the company.

Washington targets AI’s next bottleneck

The Commerce Department’s decision also reflects a change in U.S. industrial policy.

Early semiconductor support focused heavily on general fab capacity. The latest investments are becoming more targeted, with funding directed toward technologies that could constrain AI infrastructure—including photonics, advanced packaging, and power.

Stephen Empedocles, CEO of Clark Street Associates, an advisory firm specializing in securing government funding and strategic partnerships for hard tech companies, said the bottleneck in AI systems is increasingly the movement of data rather than the processor alone.

“What you’re seeing is the entire supply chain vertically integrated that they’re funding within the U.S.,” he told EE Times.

VJ Sahi, a partner at Clark Street Associates, said the award should not be viewed primarily as an attempt to catch up. “I don’t think it’s fair to say it’s addressing a market failure as much as it wants to make sure that there is a U.S. leadership position here that is maintained,” he told EE Times. “It’s a hedge to make sure we [maintain] the most advanced photonic technology.”

The award also reflects confidence in an established manufacturer with active customers and a defined production roadmap, accelerating an existing strategy rather than an unproven concept.

Geography is part of the rationale. Advanced packaging and optical-module assembly remain heavily concentrated in Asia, where some labor-intensive integration steps are less expensive. Bartlett said GlobalFoundries wants to establish U.S. capabilities spanning wafer technology, advanced materials, fiber attachment, assembly, and testing, while continuing to offer customers geographic diversity through its operations in the U.S. and Singapore.

The execution test is still ahead

The award remains a letter of intent, not a final contract, and the most important milestones have not been made public.

GlobalFoundries said it expects first-generation SCALE products to move toward volume manufacturing in 2028. The second generation should be ready for customer prototyping by the end of 2028, with production potentially beginning in 2029.

Clark Street argues that government payments should be tied to milestones that require GlobalFoundries to deliver more than it would under its existing commercial roadmap. But measuring that incremental impact would be difficult, especially because the market is already moving in GlobalFoundries’ direction.

The earnings call made clear that demand is no longer speculative.

GlobalFoundries’ data center business is growing rapidly, customers are taping out designs, and capacity is tightening in the technologies that support optical networking. The $300 million commitment is therefore not merely a subsidy for future research. It’s a geopolitical bet that GlobalFoundries can turn present demand into a higher-value, domestically anchored photonics manufacturing platform before AI infrastructure encounters its next bottleneck.

---

Read also:

CEA-Leti Pushes Stacking Roadmap as AI Runs into Memory and Power Limits

From Co-Packaged Optics to Nanolasers, Photonics Moves Inward

Intel Foundry Improves Execution, but External Customers Remain the Test

AI AND BIG DATA, DATA CENTERS, OPTICS & PHOTONICS, SEMICONDUCTORS, SILICON PHOTONICS, WAFER MANUFACTURING

GLOBAL FOUNDRIES